A Study on Autonomous Driving Motion Sickness from the Perspective of Multimodal Human Signals.

Area of Science:

• Human-computer interaction
• Automotive engineering
• Neuroscience

Background:

• Motion sickness (MS) in autonomous vehicles is a significant barrier to adoption.
• Current MS assessment relies heavily on subjective questionnaires, lacking objective quantification.
• Multimodal human signals offer potential for objective MS level (MSL) measurement.

Purpose of the Study:

• To investigate the association between multimodal human signals and MSL in autonomous driving.
• To identify key physiological and behavioral features correlating with MSL.
• To quantify the contribution of different sensor domains to MSL prediction.

Main Methods:

• Curated a dataset (HS-Set) from a decade of MS studies.
• Collected subjective MSL (fast MS scale, SSQ) and human signals (EEG, PPG, EDA, skin temp, head/eye movement) in a simulator.
• Employed correlation analysis and an explainable boosting machine (EBM) for feature and domain contribution analysis.

Main Results:

• Head kinematics (amplitude/energy) correlated with SSQ scores.
• Eye movement entropy positively correlated with nausea and oculomotor symptoms.
• Electrodermal activity (EDA) negatively correlated with nausea.
• EEG connectivity and head kinematics were dominant predictors of MSL.
• A combined Head, PPG, and EDA model retained over 80% of the full model's interpretability.

Conclusions:

• Multimodal physiological and behavioral signals can objectively quantify MSL in autonomous driving.
• EEG and head kinematics are crucial for MSL assessment.
• Combining Head, PPG, and EDA signals offers a promising, interpretable approach for MSL monitoring.